frysk builds cleanly on Fedora 7

frysk builds cleanly on Fedora 7

[Andrew Cagney]

I stepped up and fixed all the warning problems that almost everyone, as 
I predicted, by specifying the ignore-warnings-as-errors option :-(

There are still test failures which are always on the hit list.

tlrh

Re: frysk builds cleanly on Fedora 7

[Elena Zannoni]

Andrew Cagney wrote:
> I stepped up and fixed all the warning problems that almost everyone, 
> as I predicted, by specifying the ignore-warnings-as-errors option :-(
>

To clarify, the fc7 rpm itself is built with that option.


> There are still test failures which are always on the hit list.
>

Test results for fc7 have been available since June 6 on the
automated build and test system. Note that there are several test failures
on fc6 as well, including the "make dist" target.

Re: frysk builds cleanly on Fedora 7

[Kris Van Hees]

On Wed, Jun 13, 2007 at 09:36:05AM -0400, Elena Zannoni wrote:
> Andrew Cagney wrote:
> >I stepped up and fixed all the warning problems that almost everyone, 
> >as I predicted, by specifying the ignore-warnings-as-errors option :-(
> >
> 
> To clarify, the fc7 rpm itself is built with that option.
> 
> 
> >There are still test failures which are always on the hit list.
> >
> 
> Test results for fc7 have been available since June 6 on the
> automated build and test system. Note that there are several test failures
> on fc6 as well, including the "make dist" target.

Also note that as an extra feature of being able to do builds with
--disable-werror, we now have historical data that indicates that the
warnings reported on Fedora 7 were not affecting the test results.
Mind, the warnings indeed needed to get fixed, but it is very useful to
see confirmation that the warnings were not affecting functionality.

	Cheers,
	Kris

Re: frysk builds cleanly on Fedora 7

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 785 bytes --]

On Wed, 2007-06-13 at 09:36 -0400, Elena Zannoni wrote:
> Andrew Cagney wrote:
> > There are still test failures which are always on the hit list.
> >
> Test results for fc7 have been available since June 6 on the
> automated build and test system. Note that there are several test failures
> on fc6 as well, including the "make dist" target.

Currently on Fedora 7 x86 I am seeing 3 failures and 62 errors (most of
the errors seem to be cause by a "silent" failure of javah that doesn't
stop the build and that leads to some classes not being build somehow
that I haven't tracked down completely, SharedType seems missing).
On Fedora 6 on x86_64 there seem to be 2 failures and 1 error now.
Both clean cvs, both only running the frysk-core testsuite.

Cheers,

Mark

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: frysk builds cleanly on Fedora 7

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 1938 bytes --]

On Thu, 2007-06-14 at 13:57 +0200, Mark Wielaard wrote:
> On Wed, 2007-06-13 at 09:36 -0400, Elena Zannoni wrote:
> > Andrew Cagney wrote:
> > > There are still test failures which are always on the hit list.
> > >
> > Test results for fc7 have been available since June 6 on the
> > automated build and test system. Note that there are several test failures
> > on fc6 as well, including the "make dist" target.
> 
> Currently on Fedora 7 x86 I am seeing 3 failures and 62 errors (most of
> the errors seem to be cause by a "silent" failure of javah that doesn't
> stop the build and that leads to some classes not being build somehow
> that I haven't tracked down completely, SharedType seems missing).

Most of these errors seem to be instances of bug
http://sourceware.org/bugzilla/show_bug.cgi?id=4601
"Fedora 7 stacktrace problem: lib.dw.DwarfException: No class
SharedType"

While building I also get a lot of javah stacktraces in frysk.sys:
Exception in thread "main" java.io.IOException: can't find class file gnu/classpath/tools/getopt/Option.class in java.net.URLClassLoader{urls=[file:/home/mark/s
rc/frysk-obj/frysk-imports/TestPipePair.tmp/,file:/usr/lib/jvm/java-1.5.0-gcj-1.5.0.0/jre/lib/rt.jar], parent=gnu.gcj.runtime.SystemClassLoader{urls=[file:./], 
parent=gnu.gcj.runtime.ExtensionClassLoader{urls=[], parent=null}}}
   at gnu.classpath.tools.javah.Main.getClass(libgcj-tools.so.8rh)
   at gnu.classpath.tools.javah.ClassWrapper.linkSupers(libgcj-tools.so.8rh)
   at gnu.classpath.tools.javah.ClassWrapper.printFully(libgcj-tools.so.8rh)
   at gnu.classpath.tools.javah.CniIncludePrinter.printClass(libgcj-tools.so.8rh)
   at gnu.classpath.tools.javah.Main.writeHeaders(libgcj-tools.so.8rh)
   at gnu.classpath.tools.javah.Main.run(libgcj-tools.so.8rh)
   at gnu.classpath.tools.javah.GcjhMain.main(libgcj-tools.so.8rh)

These don't abort the build process though.

Cheers,

Mark

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: frysk builds cleanly on Fedora 7

[Andrew Cagney]

Mark Wielaard wrote:

Is there a bug for this one?  I suspect, though, that it has something 
to do with frysk having its own local copy of gnu/classpath/tools/getopt/.

Andrew

> While building I also get a lot of javah stacktraces in frysk.sys:
> Exception in thread "main" java.io.IOException: can't find class file gnu/classpath/tools/getopt/Option.class in java.net.URLClassLoader{urls=[file:/home/mark/s
> rc/frysk-obj/frysk-imports/TestPipePair.tmp/,file:/usr/lib/jvm/java-1.5.0-gcj-1.5.0.0/jre/lib/rt.jar], parent=gnu.gcj.runtime.SystemClassLoader{urls=[file:./], 
> parent=gnu.gcj.runtime.ExtensionClassLoader{urls=[], parent=null}}}
>    at gnu.classpath.tools.javah.Main.getClass(libgcj-tools.so.8rh)
>    at gnu.classpath.tools.javah.ClassWrapper.linkSupers(libgcj-tools.so.8rh)
>    at gnu.classpath.tools.javah.ClassWrapper.printFully(libgcj-tools.so.8rh)
>    at gnu.classpath.tools.javah.CniIncludePrinter.printClass(libgcj-tools.so.8rh)
>    at gnu.classpath.tools.javah.Main.writeHeaders(libgcj-tools.so.8rh)
>    at gnu.classpath.tools.javah.Main.run(libgcj-tools.so.8rh)
>    at gnu.classpath.tools.javah.GcjhMain.main(libgcj-tools.so.8rh)
>
> These don't abort the build process though.
>
> Cheers,
>
> Mark
>   

Re: frysk builds cleanly on Fedora 7

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 2910 bytes --]

Hi,

On Fri, 2007-06-15 at 12:58 -0400, Andrew Cagney wrote:
> Is there a bug for this one?

There is now: http://sourceware.org/bugzilla/show_bug.cgi?id=4653
Most issues were fixed by your latest patch to the common build files.
But some others still remain (see the bug report).

But this wasn't the cause of the many failures on F7. Nurding seems to
have fixed most of those with:

2007-06-15  Nurdin Premji  <npremji@redhat.com>

	* DwarfDieFactory.java (loadDies): Call System.err.println 
	for each class, avoids ecj optimizing out these classes.

Although calling println seems a little yucky. But it did seem to work.

Now Fedora 7 (x86) seems to get somewhat similar results (in frysk-core)
compared to Fedora Core 6 (x86_64) for me. Although there are still some
errors/failures on both:

Fedora Core 6 x86_64:
There was 1 error:
1) testHpdBreakpointInline(frysk.bindir.TestFhd)frysk.expunit.TimeoutException: Timeout of 5 expired
There were 4 failures:
1) testSliceAddressSpace(frysk.proc.ptrace.TestByteBuffer)junit.framework.AssertionFailedError: unexpected signal Sig_IO
2) testAttachToAttachedOtherTask(frysk.proc.TestTaskObserver)junit.framework.AssertionFailedError: event loop run explictly stopped (terminated)
3) testVarOutOfScope(frysk.rt.TestDisplayValue)junit.framework.AssertionFailedError: Variable value at first breakpoint expected:<5> but was:<0>
4) testInstructionNext(frysk.rt.TestSteppingEngine)junit.framework.AssertionFailedError: line number

Fedora 7 x86:
There were 6 errors:
1) testHpdArray(frysk.bindir.TestFhd)frysk.expunit.TimeoutException: Timeout of 5 expired
2) testHpdBreakpointInline(frysk.bindir.TestFhd)frysk.expunit.TimeoutException: Timeout of 5 expired
3) testSyscallRunning(frysk.proc.TestSyscallRunning)java.lang.NullPointerException
4) testLinuxPPC(frysk.proc.TestSyscallsWithAudit)java.lang.ArrayIndexOutOfBoundsException: 302
5) testLinuxPPC64(frysk.proc.TestSyscallsWithAudit)java.lang.ArrayIndexOutOfBoundsException: 302
6) testVarOutOfScope(frysk.rt.TestDisplayValue)java.lang.ArrayIndexOutOfBoundsException: 33
There were 4 failures:
1) testAttachToAttachedOtherTask(frysk.proc.TestTaskObserver)junit.framework.AssertionFailedError: event loop run explictly stopped (terminated)
2) testTerminateKillKILL(frysk.proc.TestTaskTerminateObserver)junit.framework.AssertionFailedError: terminating value expected:<-9> but was:<128>
3) testTerminatingKillKILL(frysk.proc.TestTaskTerminateObserver)junit.framework.AssertionFailedError: terminating value expected:<-9> but was:<128>
4) testStepOver(frysk.rt.TestSteppingEngine)junit.framework.AssertionFailedError: line number

Interestingly it seems there are some new ppc syscalls we don't
recognize according to libaudit on Fedora 7. Filed that as bug
http://sourceware.org/bugzilla/show_bug.cgi?id=4658
Anyone any ideas about the others?

Cheers,

Mark

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: frysk builds cleanly on Fedora 7

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 1303 bytes --]

On Wed, 2007-06-13 at 09:17 -0400, Andrew Cagney wrote:
> I stepped up and fixed all the warning problems

Great! It does build cleanly now on Fedora 7, but there does seem to be
a make -j dependency problem. Don't know if it is newly introduced, I
believe I have seen this before in the past. But I hit it twice already
today. It seems cni .h headers can be generated too late. Example:

../../frysk/frysk-imports/lib/dw/cni/DwException.cxx:42:32: error: lib/dw/DwException.h: No such file or directory
../../frysk/frysk-imports/lib/dw/cni/DwException.cxx:45: error: ‘lib::dw::DwException’ has not been declared
../../frysk/frysk-imports/lib/dw/cni/DwException.cxx:56: error: ‘lib::dw::DwException’ has not been declared
../../frysk/frysk-imports/lib/dw/cni/DwException.cxx: In function ‘void throwDwException()’:
../../frysk/frysk-imports/lib/dw/cni/DwException.cxx:56: error: too many arguments to function ‘void throwDwException()’
../../frysk/frysk-imports/lib/dw/cni/DwException.cxx:58: error: at this point in file
make[3]: *** [lib/dw/cni/DwException.o] Error 1
make[3]: *** Waiting for unfinished jobs....
lib/dw/DwException => lib/dw/DwException.h
lib/dw/DwarfException => lib/dw/DwarfException.h

Restarting make always fixes it though.

Cheers,

Mark

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: frysk builds cleanly on Fedora 7

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 491 bytes --]

On Thu, 2007-06-14 at 11:34 +0200, Mark Wielaard wrote:
> Great! It does build cleanly now on Fedora 7, but there does seem to be
> a make -j dependency problem. Don't know if it is newly introduced, I
> believe I have seen this before in the past. But I hit it twice already
> today. It seems cni .h headers can be generated too late.

OK, I see it isn't a new problem since Mike reported it already last
week: http://sourceware.org/bugzilla/show_bug.cgi?id=4427

Cheers,

Mark

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: frysk builds cleanly on Fedora 7

[Andrew Cagney]

The problem isn't new:

#include <lib/dw/DwarfException.h>
#include <lib/dw/DwException.h>

the build system considers that to be a system include.

Andrew


Mark Wielaard wrote:
> On Thu, 2007-06-14 at 11:34 +0200, Mark Wielaard wrote:
>   
>> Great! It does build cleanly now on Fedora 7, but there does seem to be
>> a make -j dependency problem. Don't know if it is newly introduced, I
>> believe I have seen this before in the past. But I hit it twice already
>> today. It seems cni .h headers can be generated too late.
>>     
>
> OK, I see it isn't a new problem since Mike reported it already last
> week: http://sourceware.org/bugzilla/show_bug.cgi?id=4427
>
> Cheers,
>
> Mark
>